mlx-community/Boogu-Image-0.1-Turbo-8bit

MLX (int8) conversion of Boogu-Image-0.1-Turbo (Apache-2.0) for Apple Silicon. 4-step Decoupled-DMD distilled (guidance 1.0) — ~6x faster. OmniGen2-lineage pipeline (DiT + FLUX.1 VAE + FlowMatchEuler). The Qwen3-VL-8B-Instruct text encoder is stock (bit-identical) — referenced from mlx-community/Qwen3-VL-8B-Instruct, not re-hosted.

int8 (group_size=32, attn+FFN); quant auto-detected via quant_config.json. ~12.5 GB.

Use

pip install mlx mlx-vlm && git clone https://github.com/xocialize/boogu-image-mlx && cd boogu-image-mlx && pip install -e .

from boogu_image_mlx.pipeline_mlx import BooguImagePipeline
pipe = BooguImagePipeline.from_pretrained("<this repo dir>", "mlx-community/Qwen3-VL-8B-Instruct")
img = pipe.generate("a red panda surfing on a wave, photorealistic", steps=4, guidance=1.0)  # 4-step DMD

Code: https://github.com/xocialize/boogu-image-mlx